• Proceedings of the KIEE Conference
• /
• 2003.11b
• /
• pp.167-170
• /
• 2003

본 논문에서는 패턴 인식의 전형적인 경우인 보이기 기반 물체 인식(Appearance based object recognition)을 수행하기 위하여, 일반적인 퍼지 분류 모델과, 서포트 벡터 머신을 하이브리드(hybrid) 하게 연결한 서포트 벡터 기반 퍼지 분류 시스템이라는 새로운 방법을 제안하고 이에 대하여 연구한다. 일반적인 분류(classification)문제의 경우 두 클래스로 구분하는데 최적의 성능을 가지고 있는 서포트 벡터 머신이 다중클래스(Multiclass)의 경우 발생 하는 계산량의 증가 문제를 해 결하기 위하여 다중 클래스 분류(Multiclass classification)에 장점을 가진 퍼지 분류 시스템을 도입, 서포트 벡터 머신에 연결함으로써 단점을 보완하는 시스템을 제안한다. 즉 서포트 벡터 머신을 통해 퍼지 시스템의 구조를 러닝(learning)하는데 사용하여 최종 적으로는 퍼지 분류 시스템(Fuzzy Classifier)이 나오도록 하는 것이다. 이 시스템의 성능을 확인하고자 여러 가지 물체들에 대한 이미지를 가지고 있는 COIL(Columbia Object Image Library) 데이터 베이스를 사용하여 보이기 기반 물체 인식(Appearance based Object Recognition)을 수행 하였으며 이를 순수한 서포트 벡터 머신만을 이용하여 물체 인식을 수행한 경우와 정확도 및 인식 시간에 대하여 비교하였다.

• Journal of applied mathematics & informatics
• /
• v.35 no.5_6
• /
• pp.623-649
• /
• 2017

Retrial queues have been widely used to model the many practical situations arising from telephone systems, telecommunication networks and call centers. An approximation method for a simple Markovian retrial queue by reducing the two dimensional problem to one dimensional problem was presented by Fredericks and Reisner in 1979. The method seems to be a promising approach to approximate the retrial queues with complex structure, but the method has not been attracted a lot of attention for about thirty years. In this paper, we exposit the method in detail and show the usefulness of the method by presenting the recent results for approximating the retrial queues with complex structure such as multi-server retrial queues with phase type distribution of retrial time, impatient customers with general persistent function and/or multiclass customers, etc.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2000.04a
• /
• pp.261-264
• /
• 2000

When there ate several classes of customers demanding service times with different distributions at some stations of a queueing network, the stability problem becomes suddenly complicated compared with the single class case. Recently many researchers had tried to find some kind of stability conditions for multiclass queueing networks, but did not get significant results except in very limited 2-station cases. In this study, we try to develop some dynamic control techniques which can guarantee the stability under the nominal traffic condition. Our approach includes the randomization method and the leaky bucket control scheme. Also, we mention other possibilities such as the discrete-review approach and the generalized round-robin technique. Both theoretical and experimental results will be presented.

• Management Science and Financial Engineering
• /
• v.2 no.1
• /
• pp.123-145
• /
• 1996

We consider a priority-based multiclass queue with probabilistic feed-back. There are J service stations. Each customer belongs to one of the several priority classes, and the customers of each class arrive at each station in a Poisson process. A single server serves queued customers on a priority basis with a nonpreemptive scheduling discipline. The customers who complete their services feed back to the system instantaneously and join one of the queues of the stations or depart from the system according to a given probability. In this paper, we propose a new method to simplify the analysis of these queueing systems. By the analysis of busy periods and regenerative processes, we clarify the underlying system structure, and systematically obtain the mean for the sojourn time, i.e., the time from the arrival to the departure from the system, of a customer at every station. The mean for the number of customers queued in each station at an arbitrary time is also obtained simultaneously.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06d
• /
• pp.219-221
• /
• 2012

The concave utility in the Network Utility Maximization (NUM) problem is only suitable for elastic flows. In networks with multiclass traffic, the utility can be concave, linear, step or sigmoidal. Hence, the basic NUM becomes a nonconvex optimization problem. The current approach utilizes the standard dual-based decomposition method. It does not converge in case of scarce resource. In this paper, we propose an algorithm that always converges to a local optimal solution to the nonconvex NUM after solving a series of convex approximation problems. Our techniques can be applied to any log-concave utilities.

• Journal of the Korean Data and Information Science Society
• /
• v.21 no.3
• /
• pp.561-567
• /
• 2010

Multiclassification is typically performed using voting scheme methods based on combining a set of binary classifications. In this paper we use multiclassification method with a hat matrix of least squares support vector machine (LS-SVM), which can be regarded as the revised one-against-all method. To tackle multiclass problems for large data, we use the $Nystr\ddot{o}m$ approximation and the quadratic Renyi entropy with estimation in the primal space such as used in xed size LS-SVM. For the selection of hyperparameters, generalized cross validation techniques are employed. Experimental results are then presented to indicate the performance of the proposed procedure.

• Journal of Food Hygiene and Safety
• /
• v.11 no.2
• /
• pp.89-97
• /
• 1996

This study was carried out to determine the pesticide residues in rice bran, crude rice bran oil and the oil of various stages of refining process. Each samples were analyzed for 41 pesticide residues by multiclass multiresidue methods with GC-ECD, NPD and identified by GC-MSD. Rice bran were detected cypermethrin, diazinon, dichlofluanid, and its level were ranged from 0.01~0.122 ppm. Crude rice bran oil were detected cypermethrin, diazinon, dichlofluanid, dimethoate, etrimfos, flucythrinate, and its level were ranged from 0.015~0.654 ppm Crude rice bran oil has the higher level of pesticide residues and more varieties of pesticides than rice bran. But pesticide residues in the crude rice bran oil was found to be almost removed then pigment was decolorized by absorption using active carbon and clealy removed by thermolysis for deodorization.

• Communications for Statistical Applications and Methods
• /
• v.14 no.3
• /
• pp.507-516
• /
• 2007

Multinomial logistic regression is probably the most popular representative of probabilistic discriminative classifiers for multiclass classification problems. In this paper, a kernel variant of multinomial logistic regression is proposed by combining a Newton's method with a bound optimization approach. This formulation allows us to apply highly efficient approximation methods that effectively overcomes conceptual and numerical problems of standard multiclass kernel classifiers. We also provide the approximate cross validation (ACV) method for choosing the hyperparameters which affect the performance of the proposed approach. Experimental results are then presented to indicate the performance of the proposed procedure.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.32 no.3
• /
• pp.63-80
• /
• 2007

This study is a generalization of 'stable dynamics' recently suggested by Nesterov and de Palma[29]. Stable dynamics is a new model which describes and provides a stable state of congestion in urban transportation networks. In comparison with user equilibrium model that is common in analyzing transportation networks, stable dynamics requires few parameters and is coincident with intuitions and observations on the congestion. Therefore it is expected to be an useful analysis tool for transportation planners. An equilibrium in stable dynamics needs only maximum flow in each arc and Wardrop[33] Principle. In this study, we generalize the stable dynamics into the model with multiple traffic classes. We classify the traffic into the types of vehicle such as cars, buses and trucks. Driving behaviors classified by age, sex and income-level can also be classes. We develop an equilibrium with multiple traffic classes. We can find the equilibrium by solving the well-known network problem, multicommodity minimum cost network flow problem.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2015.11a
• /
• pp.74-76
• /
• 2015

본 논문에서는 효율적인 자세인식을 위해 인체 골격 정보를 활용한 멀티클래스 SVM(Multiclass Support Vector Machine)학습 기반의 자세 인식 분류 기법을 제안한다. RGB 카메라로 취득한 영상을 활용하거나 깊이 카메라로부터 취득한 골격 정보를 그대로 사용하는 기존 연구와 달리 제안 기법에서는 깊이 정보로부터 추출한 인체의 3 차원 골격 정보를 이용하여 고차원의 특징을 추출하고 그로부터 자세 인식 분류를 수행한다. 제안 기법의 특징 벡터는 깊이 정보에서 취득한 골격 정보의 관절간 각도의 조합으로 구성하여 인체의 골격 편차에 강인할 뿐 아니라 특징의 차원을 효과적으로 감소시킬 수 있다. 또한 분류기로는 멀티클래스 SVM 방식 중 one-vs-one 분류 방식을 이용하여 학습 및 판별을 수행함으로써 제안 기술의 성능을 평가한다. 실험을 통해 제안 기법은 다수의 자세에서 비교하는 다른 학습 기법보다 비교적 높은 자세인식률을 보인다.